Motivation: we want to reinstate D24481 / rL282259 without breaking test-suite bots, so let's generalize the check that PowerPC is currently using to overcome the too-strict reference output checks. Ideally, we would do something like rL282027 to customize the checking per test, but let's not gate FMA codegen on that because it requires more investigation to get right.

From the earlier failures, we know that future improvements in this area will require changes for output-checking for at least these tests when running in a mode where fp-contraction is allowed:

FAIL: MultiSource/Applications/oggenc/oggenc.execution_time (490 of 2445)

FAIL: MultiSource/Benchmarks/MiBench/telecomm-FFT/telecomm-fft.execution_time (491 of 2445)

FAIL: MultiSource/Benchmarks/VersaBench/beamformer/beamformer.execution_time (492 of 2445)

FAIL: SingleSource/Benchmarks/Linpack/linpack-pc.execution_time (493 of 2445)

FAIL: SingleSource/Benchmarks/Misc-C++/Large/sphereflake.execution_time (494 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/datamining/correlation/correlation.execution_time (495 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/datamining/covariance/covariance.execution_time (496 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/2mm/2mm.execution_time (497 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/3mm/3mm.execution_time (498 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/atax/atax.execution_time (499 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/bicg/bicg.execution_time (500 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/cholesky/cholesky.execution_time (501 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/gemver/gemver.execution_time (502 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/gesummv/gesummv.execution_time (503 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/mvt/mvt.execution_time (504 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/symm/symm.execution_time (505 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/trisolv/trisolv.execution_time (506 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/kernels/trmm/trmm.execution_time (507 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/linear-algebra/solvers/gramschmidt/gramschmidt.execution_time (508 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/stencils/adi/adi.execution_time (509 of 2445)

FAIL: SingleSource/Benchmarks/Polybench/stencils/fdtd-2d/fdtd-2d.execution_time (510 of 2445)